Professional Documents
Culture Documents
2017 - Ippoliti Methods and Finance A View From Outside
2017 - Ippoliti Methods and Finance A View From Outside
2017 - Ippoliti Methods and Finance A View From Outside
from Outside
Emiliano Ippoliti
Abstract The view from outside on finance maintains that we can make sense of,
and profit from, stock markets’ behavior, or at least few crucial properties of it, by
crunching numbers and looking for patterns and regularities in certain sets of data.
The basic idea is that there are general properties and behavior of stock markets that
can be detected and studied through mathematical lens, and they do not depend so
much on contextual or domain-specific factors. In this sense the financial systems
can be studied and approached at different scales, since it is virtually impossible to
produce all the equations describing at a micro level all the objects of the system
and their relations. The typical view of the externalist approach is the one provided,
for instance, by the application of statistical physics. By focusing on collective
behaviors, statistical physics neglects all the conceptual and mathematical intrica-
cies deriving from a detailed account of the inner, individual, and at micro level
functions of a system. This chapter examines how the view from outside deals with
critical issues such as the mathematical modeling (Sect. 2), the construction and
interpretation of data (Sect. 3), and the problem of prediction and performativity
(Sect. 4).
1 An Overview
Two views shape our understanding and approaches to finance, and stock markets
in particular—as well as to a lot of other domains of human knowledge: the view
from inside and the view from outside. They employ radically different assumptions
about the set of the relevant features of the domain and the way to look at them—
e.g. the ontology, the methods, and the role of mathematical modeling. The book
examines these two views, analyzing their interconnections—when and where they
clash, reconcile, or act independently—and the ways by which these views shape
various approaches to the study of finance and its branches.
E. Ippoliti (✉)
Department of Philosophy, Sapienza University of Rome, Rome, Italy
e-mail: emi.ippoliti@gmail.com
emi.ippoliti@gmail.com
4 E. Ippoliti
Accordingly, these views tackle in a very different fashion the two main sides of
the problem of stock markets prices behavior—namely the quantitative and the
qualitative side. The quantitative side requires figuring out an answer to prices’
series—that is a mathematical tool to approximate them in the most cogent way.
The qualitative side requires determining the variables that affect the dynamics of
stock market prices, and their “machinery”, as the result of the actions, or better
interaction, of investors who sell and buy stocks. The two sides, of course, are not
totally separated and they interact in several ways on different problems.
In this chapter I will focus on the view from outside, or the externalist view.
The externalist view argues that we can make sense of, and profit from stock
markets’ behavior, or at least few crucial properties of it, by crunching numbers and
looking for patterns and regularities in certain sets of data. The notion of data,
hence, is a key element in such an understanding and the quantitative side of the
problem is prominent—even if it does not mean that a qualitative analysis is
ignored. As a matter of fact, few externalist approaches try to put together a
qualitative analysis along with a quantitative one. But the point here that the outside
view maintains that it provides a better understanding than the internalist view. To
this end, it endorses a functional perspective on finance and stock markets in
particular (see Bodie-Merton [3], p. 24–32). The focus on functions rather than
entities or institutions seems to offer more stable objects of inquiry, from which the
formers can be derived.
The basic idea of the externalist view is that there are general properties and
behavior of stock markets that can be detected and studied through mathematical
lens, and they do not depend so much on contextual or domain-specific factors. The
point at stake here is that the financial systems can be studied and approached at
different scales, and it is virtually impossible to produce all the equations describing
at a micro level all the objects of the system and their relations. So, in response, this
view focuses on those properties that allow us to get an understanding of the
behavior of the systems at a global level without having to produce a detailed
conceptual and mathematical account of the inner ‘machinery’ of the system. Here
two roads open. The first one is to embrace an emergentist view (see e.g. Bedau—
Humphreys [2, 21]) on stock market, that is a specific metaphysical, ontological,
and methodological thesis: in this case it would be even useless to search for and
write down all the equations for all the objects of the system and their relations at
the micro level, since the properties at a different scale are not reducible to the one
of the scales below it. The second one is to embrace a heuristic view, that is the idea
that the choice to focus on those properties that are tractable by the mathematical
models is a pure problem-solving option. This does not imply that we cannot
account for the global behavior of the system in terms of its micro level dynamics
and constituents, in the future. Maybe it requires a better understanding of the micro
behavior of the systems, further data, and, in case, new pieces of mathematics.
A typical view of the externalist approach is the one provided, for instance, by
statistical physics. In describing collective behavior, this discipline neglects all the
conceptual and mathematical intricacies deriving from a detailed account of the
inner, individual, and at micro level functioning of a system. Concepts such as
emi.ippoliti@gmail.com
Methods and Finance: A View from Outside 5
emi.ippoliti@gmail.com
6 E. Ippoliti
scales of their posterior evolution. Such extreme events express more than anything else the
underlying “forces” usually hidden by almost perfect balance and thus provide the potential
for a better scientific understanding of complex systems [21], 17–18).
emi.ippoliti@gmail.com
Methods and Finance: A View from Outside 7
action) all information contained in the market prices, we propose that the market as a
whole can exhibit “emergent” behavior not shared by any of its constituents [21], 279).
2 Mathematical Modeling
Mathematical modeling1 plays a pivotal role in the view from outside, not only
because it aims at answering the quantitative side of the problem of stock market
prices, but especially because it maintains that it is just through the quantitative side
of the problem that we can eventually find a solution to the general problem. By
solution here we mean a way to foresee future trends in stock markets, or at least
precursory signs of critical events—i.e. crashes. Accordingly, the choice of the
1
See [16] for a detailed account for the role of models in economics.
emi.ippoliti@gmail.com
8 E. Ippoliti
emi.ippoliti@gmail.com
Methods and Finance: A View from Outside 9
On one hand, the mathematical tools usually employed in financial analysis are
basically the same as those used in economic analysis, on the other hand financial
data, more often than not, are different from the macroeconomic data (w.r.t. fre-
quency, seasonality, accuracy, etc.). While in economics we have problems such as
the small samples problem, measurement error or data revisions, in finance they
are much, much less common, if any.
In the first case, the small samples problem, for instance economic findings for
government budget deficits, data are measured only on an annual basis, and if the
technique used to measure these variables changed, let us say, a 25 year ago, then
only twenty-five of these annual observations would be useful (at most). In the
second and third case, the data might be estimated or measured with error, leading
to subsequent revisions. In general these problems are rare in finance. In effect
financial data are so that:
• prices and other variables are recorded as a transaction takes place, or when are
quoted on the displays of information providers.
• some sets of financial data are recorded at much higher frequencies than
macroeconomic data—at daily, hourly, and now even
milliseconds-by-milliseconds frequencies. Here we enter the realm of ‘big data’
and the number of findings available for analysis can potentially be very large.
This means that powerful techniques can more easily be applied to financial
rather than to economic data, and the results are more reliable.
Moreover, financial data are not regular (in time): the price of common stocks
for a company, for instance, generally is recorded whenever there is a new trade or
quotation placed by the financial information recorder. Such recordings are very
unlikely to be evenly distributed over time.3 A standard way of dealing with this
issue is to take a certain frequency and use the last prevailing price during the
interval of it. Moreover a standard way of treating financial data is to employ all
2
The equation, developed by Sornette, can be expressed in the following simple way: y(t) =
A + B (tc − t) z + C (tc − t) z cos (ω log (tc − t) + Φ).
3
Usually there is no activity when the market is closed, there is less activity around the opening
and closing of the market, and around lunch time.
emi.ippoliti@gmail.com
10 E. Ippoliti
data of the same frequency. This means that if we are trying to estimate, for
instance, a model for stock prices using weekly observations on macroeconomic
variables, then we must also use weekly observations on stock returns, even if
observations on the latter variable are recorded monthly.
Furthermore, the financial data exhibits other problems, as their ‘noisy’—so that
it is more complicated to draws apart underlying trends or patterns on one hand and
random or uninteresting features on the other. Furthermore, the set of high fre-
quency data often presents additional properties and dynamics, which are a con-
sequence of the way the market works, or the way the prices are recorded.
In sum, the understanding of the notion of data, and their construction, in
financial systems and stock markets in particular, is essential. Just to give an
example of the complexity of financial data let us consider a much-talked piece of
financial data, that is the alleged dimension of financial markets. How big are
financial markets? A common figure, calculated on data provided by the Fed and
World Bank, says us that the financial markets are very large, maybe too large: they
are much larger than the economy—a multiple of it! For instance we know that in
2011:
(1) the total amount of debt in US was 54,2 trillions dollars (loans only—source:
FED)
(2) the US stock market was 15,6 trillions dollars (source: World Bank)
(3) the US money supply was 10 trillions dollars (source: FED)
So, adding (1) and (2) we get a total about 70 trillions dollars as a good esti-
mation of the dimension of financial markets. Moreover we know that:
(4) the GDP (Gross Domestic Product), that is the total value of the financial goods
and services produced in an economy in one year (a benchmark for measuring
the economy of a country), was about 15 trillions in 2011 for US.
So we have that US stock market (2) alone is more ore less equal to Us GDP (4)!
And also that a great portion of US financial markets (1 + 2) is five times US GDP
(4)!
Why do the stock markets seem so large? The answer is in the very construction
of the data. In fact, here we are merging in a single piece of data (a number) two
variables of different kind, namely stock (a) and flow (b) variables:
(a) the size of financial markets, an index of wealth, which measure the current,
static value of all assets we own now, how much we have of something at a
point in time.
(b) GDP, an index of income, which measures an activity, the money we earn each
month/year, and hence how much we did something over a specific period of
time.
The point here is that variable (a) has the problem of so-called double or multiple
counting, that is the fact that a transaction is counted more than once. Money travels
emi.ippoliti@gmail.com
Methods and Finance: A View from Outside 11
trough the financial markets: an income of one institutional unit is the expenditure of
another, and what is an input of one institutional unit can be the output of another.
This passage, which often can be more than one, affects our data on the size of
the financial markets. Let us consider for instance a typical situation: a household
making a deposit in a local bank, say 10.000 €. This seemingly simple act can
generate, and usually does, several assets. The local bank can make a loan to a
regional bank, and this, in turn, lends money to a corporation that undertakes a big
project. In this case the household deposit has passed along twice, i.e. three busi-
ness have used these money to generate assets: each bank has a new asset (a loan),
and the corporation has a new asset (a new project). In that case also new debts are
generated: the one that the local bank owns to the depositor, the one the regional
bank owns to the local bank, and the one the corporation owns to the regional bank.
So the same amount of money, 10.000 €, generates a lot of multiple counting, that
are recorded in the statistics and our data about financial markets. That explains
why they seem so big: the more complex the path that money takes to get from the
original lender to the ultimate borrower, the bigger the financial system will appear
to be. Thus not only it is no surprising that financial markets are bigger then the
economy, but in a sense they should be so—and this can also be good. They are
really big and this means that the financial system (1) has backups, and (2) the there
is more competition between the lenders, a fact that lowers the costs and improves
the quality of the financial services.
Before pointing at a crucial property of stock markets data in particular, it is
worth recalling here that there are basically three kinds of data in quantitative
analysis of financial phenomena: time series data, cross-sectional data, and panel
data.
Time series data are collected over a period of time for one or more variables.
More precisely time series data are built over a particular frequency of observation
—or data points. The frequency is simply a measure of the interval over which the
data are recorded. For instances data about industrial production are recorded
monthly or quarterly, while government budget deficit on annual base. The data
may be quantitative (e.g. prices or number of shares outstanding), or qualitative
(e.g. the day of the week or a credit rating, etc.). These data are employed to mine
correlations where time is the most important dimension. For instance, they can be
used to examine how the value of a stock index of a country changes as the
macroeconomic fundamentals of it change, or how the value of a company’s stock
price has changed after the announcement of the value of its dividend payment.
At the other side of the temporal scale we find the cross-sectional data, that is
data on one or more variables collected at a single point in time (hence without
differences in time). These data offer an instant picture of a system and tell nothing
about the history or the changes of it over time—for example, the data might be a
sample of bond credit ratings for USA banks.
This kind of data are employed to tackle problems like the relationship between
company size and the return to investing in its shares, the relationship between a
country’s GDP level or the probability that the government will default on its
sovereign debt.
emi.ippoliti@gmail.com
12 E. Ippoliti
Then we find the panel data, which have the dimensions of both time series and
cross-sections, e.g. the daily prices of a number of blue chip stocks over two years.
Thus panel data provide observations of multiple phenomena over multiple time
periods for the same entities (e.g. firms or individual). In this sense, time series and
cross-sectional data can be seen as panel data of just one dimension.
We are now in a position to examine a crucial property of the data about stock
markets. In effect these data come in a very peculiar form: we do not know who
trades what. We have records for time, volume, price of stocks, but we miss this
fundamental piece of information: trading-account identifiers are nearly always cut
out and this makes almost impossible to detect sequences of actions by the same
executing trader. This could be a serious obstacle in the understanding and mod-
eling of stock markets and financial systems. How can we read the dynamics of a
social system if we do non know who is doing what?
Here the view from outside seems to score a point. As a matter of fact, the absence
of this kind of data is not a great concern for an externalist view: if this approach is
right, you can ignore this piece of data to a large extent without compromising the
understanding of crucial proprieties and dynamics of stock markets. Since it is the
collective behavior that counts and will determine the overall dynamics of the system,
the specific information of the individual behavior and choices can be safely ignored.
What is possible at the individual level it is not possible at an aggregate level and vice
versa. The peculiarities and possible deviances of a single individual’s behavior are
cancelled out and “absorbed” in the collective behavior.
It is not coincidence that a paradigmatic externalist approach, like the one put
forward by Sornette, draws on an analogy with geology, and seismology in par-
ticular. Just as in seismology, where most events occur at a depth such that there is
no way to directly measure the force (the stress), in stock markets we do not have a
way (the data) of seeing the force acting in depth and at an individual level. We
have to measure it indirectly, statistically, and collectively.
The point here is to what extent and under what conditions this piece of
information can be considered as negligible for the understanding of stocks mar-
kets. The view from outside takes a clear-cut position, but the problem still remains
open, at least in part. Maybe you can ignore this piece of data in the study of certain
few dynamics of the stock markets, but in other cases you should consider this very
carefully, especially in case of asymmetries, for instance when few actors are
capable to affect the system or large part of it.
The outside view goes under a stress test with the notion of prediction, which becomes
so controversial in stock markets. One of the main theoretical problems in finance, and
in the social sciences in general, as argued by Flanagan [10] and then by practitioners
like George Soros [23], is its ‘reflexivity’: the fact that a prediction, and hence an
expectation, about the behavior of the system is able to affect and change it. This is
emi.ippoliti@gmail.com
Methods and Finance: A View from Outside 13
arguably one of the greatest differences between social sciences, where systems are
able to learn and are reflexive, and natural sciences, where systems do not exhibit such
a feature. This fact makes it very difficult to evaluate a financial hypothesis by means
of a comparison between a prediction about it and the events that happen in reality, as
the former can affect the latter. As an example, consider a prediction released into a
financial system (e.g. the forecast that a crash of about 15–25% will occur in six
weeks). It will generate one of the following three scenarios depending on the
threshold of how many investors believe the prediction:
(a) not enough investors believe it, which then is useless, and the market drops just
as predicted. Even if this seems a victory for the hypothesis that generates the
prediction, criticism can define it as ‘lucky one’, which does not have any
statistical significance.
(b) A sufficient number of investors believe it, who adjust their strategies accord-
ingly and the bubble vanishes: the crash does not occur and the forecast is
self-refuting.
(c) A lot of investors believe it, causing panic, and the market drops as a conse-
quence. So the forecast is self-fulfilling and its success is due to the effect of the
panic rather than to the predictive power of the hypothesis.
These scenarios are problematic for the theory that forecast them: in case (a) and
(c) the crash is not avoided, while in (b) the forecast is self-refuting and accordingly
the theory turns out to be unreliable. This feature seems to be a defining property of
the learning and reflexive systems and raises also the problem of scientific
responsibility, in particular the responsibility to publish scientific findings.4
This feature of the stock markets has been re-conceptualized and refined by
Callon [5] and MacKenzie [11], who examined the notion of performativity in stock
markets. In particular MacKenzie examined the ways in which the financial models
and theories influence and shape the systems they seek to understand. These
models, quoting Mackenzie, are an engine and not a camera of financial markets:
“financial economics, I argue, did more than analyze markets; it altered them. It was
an ‘engine’ in a sense not intended by Friedman: an active force transforming its
environment, not a camera passively recording it” (MacKenzie [11], 12).
Now, performativity can be broken down in two parts. On one hand we find the
so-called Barnesian performativity. It describes those situations where the practical
employment of an economic theory makes economic phenomena resemble what
they are described to be by that economic theory. On the other hand we find the
so-called counter-performativity, where the employment of an economic model
makes economic phenomena differ from the description of them given by this
model. Nowadays, quantitative trading (QT) is done by most of the major players in
the financial markets, especially the ones committed to an externalist view. QT of
4
This issue becomes much more complex when researchers have to take into account the potential
effect of the publication of their in society. The stock markets exemplify this problem and its
complexity.
emi.ippoliti@gmail.com
14 E. Ippoliti
References
1. Abergel, F., Aoyama, H., Chakrabarti, B.K., Chakraborti, A., Ghosh, A.: Econophysics of
Agent-Based Models, London: Springer (2014)
2. Bedau, M., Humphreys, P. (eds.): Emergence: Contemporary Readings in Philosophy and
Science. Cambridge (Massachusetts): The MIT Press (2008)
3. Bodie, Z., Merton, R. Finance. Prentice-Hall, New Jersey (2000)
4. Brooks, C.: Introductory Econometrics for Finance. Cambridge University Press, New York
(2008)
5. Callon, M.: Introduction: the Embeddedness of Economic Markets in Economics. In: Callon,
M. (ed.) The Laws of the Markets. Blackwell, Oxford (1998)
6. Chatterjee, A., Chakraborti, A., Chakrabarti, B.K.: Econophysics and Sociophysics: Trends
and Perspectives. Wiley-VCH, Berlin (2006)
emi.ippoliti@gmail.com
Methods and Finance: A View from Outside 15
emi.ippoliti@gmail.com
Emiliano Ippoliti ⋅ Ping Chen
Editors
123
emi.ippoliti@gmail.com
Editors
Emiliano Ippoliti Ping Chen
Department of Philosophy National School of Development
Sapienza University of Rome Peking University
Rome Beijing
Italy China
emi.ippoliti@gmail.com